THE SHORT ANSWER
Correlation means two variables vary together. Causation means a change in one contributes to producing a change in the other. Correlation alone cannot establish direction or rule out coincidence, reverse causality, confounding, selection effects or measurement problems.
Ask what produced the observed pattern
| Observation | What remains possible |
|---|---|
| Ad spend and sales rose together | Ads contributed; demand drove both; seasonality or another channel changed |
| Tool users finish faster | The tool helped; faster workers adopted it; tasks differed |
| More support tickets accompany churn | Problems caused churn; risky customers contact support more; both share another cause |
Test three common alternatives
- Confounding: another factor influences both variables.
- Reverse causality: the proposed outcome influences the proposed cause.
- Coincidence or unstable measurement: the pattern may not persist or may reflect noise.
These possibilities do not prove the causal story false. They show why the association alone is insufficient.
Evidence & context: Federal Judicial Center and National Academies of Sciences, Engineering, and Medicine
Design strengthens or weakens a causal claim
Randomized experiments can make groups more comparable and help isolate an intervention's effect. They still require sound measurement, enough data and careful interpretation. When experiments are infeasible, causal reasoning may combine study design, timing, mechanisms, comparison groups and multiple forms of evidence.
For marketing applications, continue to A/B Testing and Marketing Experimentation.
Before saying caused
- Did the proposed cause occur before the outcome?
- What else changed?
- Could the relationship run in reverse?
- How were cases selected and measured?
- What comparison or experiment would distinguish explanations?
Sources & further reading
- Reference Guide on Epidemiology
Federal Judicial Center and National Academies of Sciences, Engineering, and Medicine. A current methodological reference on association, bias, confounding and causal inference. OpenSkool uses only introductory reasoning concepts, not health conclusions.
- Methods for Measuring Brand Lift of Online Ads
Google Research. Original research using randomised experiments to estimate advertising effects; no universal lift or ROI benchmark is inferred.
Examples and exercises are illustrative unless attributed to a source. No independent expert review is claimed.
A correction, a counterexample or an experience worth sharing?
Join the conversation ↗